Research And Implementation Of HMM Based Uyghur Speech Synthesis System

There are new higher requirements to the technology of the speech synthesis along with continuous development of speech synthesis technology. The speech synthesis was required to meet with the needs of the diversification of synthetic speech; this means that speech synthesis should be able to develop synthesis system quickly using limited speech resources within a short time. Due to satisfy these needs and research purpose, this paper studied the HMM based speech synthesis and applied its principle to the Uyghur language and construct the HMM based TTS system for Uyghur language. The data preparation such as text corpus design, recording, labeling, basic units confirmation, design of context dependant attributes, design of question set, label file with time labels and label files without time labels and training part have been done successfully. The synthesize part have been done under Linux environment and synthesized speech basically normal. Improvements on synthesize speech such as duration of the last phoneme and revising few surd frames in continuous sonant was done successfully by modifying the speech parameter generation algorithm.The synthesizing part of the HMM based Uyghur speech synthesis was written under Redhat enterprise linux 5.0. Rhythm information obtained from the text to be synthesized couldn't be segmented automatically. So there has no complete text analyzer for HMM based Uyghur speech synthesis system and there has no synthesis interface under windows for HMM based Uyghur speech synthesis system. But this paper confirmed the constructing process and technical way of HMM based Uyghur speech synthesis system and the complete HMM based speech synthesis system will be developed in the near future.